For the more than 100 drugs that the US National Cancer Institute's pediatric programs have tested in mice carrying childhood tumours since 2004, how well did the mouse results predict which drugs later helped or failed in children's trials, and which features of the mouse test (share of tumours that shrank, whether the mouse drug exposure was reachable in children, drug class) carry the signal?
- Good it would do
- 2 / 5
- Can AI help
- 2 / 5
- How neglected
- 3 / 5
- Can we check it
- 3 / 5
Why it's on hold
Parked: no drug-by-drug, whole-dataset comparison of PPTP/PPTC mouse results with children's trial outcomes exists (the 2024 review's tables are mouse-only; the outside 'fatal' was not upheld), but likely fewer than 10 clinical successes, selection bias in which drugs reached children, and no exposure data in the public spreadsheet put the predictive question out of reach; only a small descriptive table is feasible, and the NCI PIVOT team holds the data. The outside skeptic calls this unfinished vetting because the cheap census that would decide it was not run.
Why it's stuck
The testing programmes (PPTP 2004 to 2014, then PPTC, now PIVOT) published results agent by agent, and their own 2024 review of 20 years ('Lessons learned from 20 years of preclinical testing in pediatric cancers', Smith, Houghton, Lock and others, Pharmacology & Therapeutics) compares mouse and clinical results only drug class by drug class, in prose. It notes that mouse results over-predicted clinical activity for antibody-drug conjugates and that the models 'will inherently overpredict the activity of agents that are active against cells only during mitosis'.
What we'd try
Join the public Zenodo summary of in-vivo response metrics (record 13871579, 'Preclinical Pediatric In Vivo Testing Data Summaries') with the outcome of every later pediatric phase 1 or 2 trial of the same drug in the same tumour type. The outcomes would be found by large-scale, model-assisted reading of ClinicalTrials.gov results and published trial reports, with a second independent extraction pass to check it.
We'd look again if
Either (1) a dated census of the ~140 agents in Zenodo sheet ST1 against ClinicalTrials.gov and PubMed finds at least about 40 drug-tumour pairs with comparable single-agent pediatric outcomes and at least about 10 clinical successes (then re-vet a narrowed descriptive concordance study, limited to agreement within clinically tested drugs); or (2) a PIVOT or Children's Oncology Group investigator says they would use such a table to rank agents and confirms the linkage has not been built internally, ideally offering mouse exposure data.
Think we got it wrong?
Ruled-out problems can come back. Tell us what we missed.
Members can read the full record for this problem, check its sources and comment. Open the member view →